Muse Wonder: How is Meta Competing at the Frontier?
Meta's new Muse Wonder models challenge OpenAI and Anthropic at the AI frontier with state-of-the-art performance on personal and professional agentic task
TL;DR
- Meta released Muse Wonder and Muse Wonder Full, its most capable AI models.
- Muse Wonder Full scores 56 on the Artificial Analysis Intelligence Index, outperforming OpenAI's GPT-6 Astra.
- The models excel at personal and professional agentic tasks, including legal research and scientific reasoning.
- Muse Wonder Full processes text, audio, images, and video with a one million token context window.
- Meta claims zero agentic misalignment and reduced catastrophic domain threats in Muse Wonder Full.
- The models are cheaper than advanced Anthropic iterations but more expensive than GPT-6 Astra or Gemini 4 Argon.
- Meta's AI efforts have undergone an overhaul in the last fourteen months, focusing on building personal AI agents.
- Key partners include NVIDIA for hardware, Microsoft for enterprise AI deployment, and EssilorLuxottica for smart eyewear.